Post on 31-Oct-2014
description
Better LivingthroughLinking
Daniel Chudnov - 2009-05-27 - dchud at umich eduTexas Conference on Digital Libraries 2009
MY OPINIONSDO NOT REFLECT
THOSE OF:MY EMPLOYERTHE US GOV’T
PRES. BARACK OBAMATHE DETROIT RED WINGS
•doing “web” better
•linking data
•steps you can take today
1.
doing “web” better
we’ve been buildinga web
for a long time
Photo cc-sa by osunick on flickr, http://tinyurl.com/oy8qw942
which brings usto about, say,
1996.
xml1997
db-backed sites1998
xml-backed sites1999
blog tools2001
better search2002
rssopenurloai-pmhsrw/u
opensearchmicroformats
redesignredesignredesignredesignredesignredesign
scriptsincludes
templatesframeworks
cmsseo
User. Generated. Content.
OMG ROFLMAO LOLZ
how we use the webevolves
how we build the webevolves
it’s not done yet,but
we are stillbuilding a web
here’s what’s next
2.
linking data
links matter
just ask Google
but we knew that
for example
which, basically,was this
the webis made of links
but waitthere’s more
which, basically,was this
butnobody knows that
...unless theyknow to look...
...which is theproblem.
more to the point
Google won’t find this.
http://authorities.loc.gov/cgi-bin/Pwebrecon.cgi?AuthRecID=2852966&v1=1&HC=3&SEQ=2009052703254
5&PID=XGImKyGgg1Cofw8uO3IFxAvS
what’s wrongwith this picture?
http://authorities.loc.gov/cgi-bin/Pwebrecon.cgi?AuthRecID=2852966&v1=1&HC=3&SEQ=2009052703254
5&PID=XGImKyGgg1Cofw8uO3IFxAvS
system info
session info
unknowableinfo
unbookmarkable
screenscraping required
unscriptable
item record authority record
not linked
a holdings recorddoesn’t help
but wait, there’s more
not connected
not findable
not relevant
ouch.
despite our efforts
the expense ofdigitizing
care and feeding of
cranky coders
the overheadof
authority control
we count onGoogle to guess
at relevance
inside our systemswe tweak
and improve
across our systemswe can
only hope
time todo
better
enterLinked Data
“Linked Data”is
data in the web
“Linked Data”is
a movement
“Linked Data”is
dogma
“Linked Data”is
worth a close look
so what is it?
a way tolink
better
a way tomake
concepts explicit
a way toconnectour stuff
item record authority record
a way toconnect
these together
draw our stuffdeeper
into the web
not justfiles to download
but part of the web
how?
use URIs as names for thingsuse HTTP URIs
provide useful informationinclude links to other URIs
http://www.w3.org/DesignIssues/LinkedData.html
break down conceptsmake them “pages”
the old
the new
the old
bad URL
screenscrapingrequired
few links out
no links in
the newgood URL
alternateformats
some links out
more links in
http://lccn.loc.gov/84759993
LCCN Permalink
alternateformats
more links out
clean URI
discoverable, parseablealternate formats
apply these ruleseverywhere
resource URI
rel=alternate
rel=contents
rel=alternate
(already has)resource URI
resource URI
resource URI
arrowsbecome
cross-references
why?
it becomescrawl-ablemine-able
integrated accessthrough
smart crawl / index
precision improveswith
rich metadataexplicitly linked
recall improveswith
explicit linksthrough authority recs
bringprecision / recallof clunky OPACs
to open web
this isdoing web stuff
better
just by doingHTML / HTTP
use URIs as names for thingsuse HTTP URIs
provide useful informationinclude links to other URIs
http://www.w3.org/DesignIssues/LinkedData.html
doesn’t fit the web,it is the web
many examples already
LIBRISid.loc.gov
chronicling america
if i squint just right...
the “LOD cloud”
a“ready reference”
collection
make this
useful
turn thisinto
linked data
seems obvious,right?
3.
steps you can take today
how tolink your data
(or, About All ThatOther Stuff You Have)
use URIs as names for thingsuse HTTP URIs
provide useful informationinclude links to other URIs
http://www.w3.org/DesignIssues/LinkedData.html
first:clean, cool URIs
for data concepts
instead of this:/search?subject=Strayhorn,+Billy
do this:/subject/lccn/n81072976
easy to implementtoday
easy to redirecttomorrow
next:offer data views
use <a rel=___ href=___>
<link rel=___>
offer upMARC,
or JSON,or RDF,
or ...
next:relate your datato established
authorities
your search resultsbacked by sharedauthority records
put it all together
visit/subject/lccn/n81072976
• get your content results
• get alternate data views (concept record)
• get links to other useful data
• get links to authorized form
all that’s easy
i promise
the next part’shard
how to make it last
if your site breakswhen links break
cache and link yourself
instead of just this
do this
UChicagocache
same as
just likewe do
in ILS/OPACs
but, yknow,the web way
so if a remotelink breaks
your local linksstill work
also, proxy concepts
your/subject/lccn/n81072976
==my
/subject/lccn/n81072976
but in your app
make every cacheits own
linked data source
instead of just this
UChicagocache
same as
enable this
UChicagocache
same as
same as
same as
same as
same as
if one goes downthe others live on
if others live onit’s all still linked
if it’s all still linkedwe can keep adding
linked datathat lasts
i think
summary
•making the web better
• follow Linked Data practices
•easy changes to make now
summary
•get more value from our data
• get more users to our data
•make relevance explicit
thanks!
dchud at umich eduonebiglibrary.net
twitter.com/dchud